Several decades of tobacco research and control in the U.S. have left a large amount of cross-sectional survey data at the national and the local levels. Typical data sources include the Youth Risk Behavioral Survey since 1975, the Behavioral Risk Factor Survey since 1984, the National Health and Nutrition Examination Survey since 1971, the National Health Interview Survey since 1957, the Monitoring the Future Studies since 1975, the National Survey on Drug Use and Health since 1971, and the California Youth Tobacco Survey. These multisource multi-year data with millions of randomly sampled subjects consist of an invaluable asset and further exploration of the data would benefit tobacco control to reach the objectives set forth by the Healthy People 2010. Effective planning and evaluation of tobacco control programs will benefit from data on smoking behavior change (e.g., from nonsmokers to smokers, from smokers to quitters and from quitters backward to smokers again), because such data are more informative than the conventionally used prevalence rates of lifetime smokers or 30-day smokers. Measuring smoking behavior change requires longitudinal data and collecting longitudinal data at the state and local levels year by year are often too costly and too complex. Our analyses indicate that cross-sectional data contain the needed longitudinal information;but a challenge remaining is how to extract the information out of the data. Through a long-term of collaboration, the investigators of this project have recognized that the Probability Discrete Event System (PDES) from the engineering and system control field can be adapted to overcome this challenge. Following Shannon's sampling principles, the PDES technique has been successfully used in modeling sophisticated machine-human interactions with cross-sectional data. The goal of this study is to adapt this method for modeling smoking behavior. The four aims of this research are: (1) to establish a PDES smoking behavior model capable of assessing smoking behavior progression with cost-saving cross-sectional data;(2) to analyze effect of common tobacco control programs (e.g., taxation, restriction of sales and use of tobacco, school-based programs and tobacco cessation) on smoking behavior progression;(3) to predict smoking trends and to simulate different tobacco control scenarios that are directly related to the tobacco control objectives set by the Healthy People 2010;and (4) to develop software that computerizes the PDES method for tobacco research and tobacco control practice. Ten-year data (1995-2004) from the National Survey on Drug Use and Health will be used to test the proposed method. The method will then be further validated with longitudinal data (e.g. National Longitudinal Survey of Youth 1997). A collaborative research team with expertise in tobacco research and system control has been established is well positioned to carry out this project. This proposed research is completely relevant to public health, because the expected results will provide needed techniques and data for advancing the existing tobacco research and strengthening the tobacco control strategies in the nation.